Help

[anish katwal]

Hi everyone,

I’m working on a SWAT model for a small lake watershed. The lake is represented as a reservoir, and my main project goal is to estimate phosphorus loading from spatially defined septic HRUs.

The watershed does not have a direct discharge gauge at the lake inlet/outlet. For hydrology calibration, I have been using discharge from a nearby gauged river watershed as a proxy, scaled by drainage area. The proxy watershed is more than four times larger than my modeled watershed and has a more river-dominated drainage system.

I calibrated against reservoir inflow (`FLOW_INcms` from `output.rsv`) because my goal is watershed inflow to the lake/reservoir. I tried raw monthly flow, then 3-month smoothed monthly flow for both observed proxy and simulated flow. Even with broad hydrologic parameter ranges and many calibration runs, the results are still poor in terms of timing:

- p-factor around `0.45-0.49`
- r-factor around `1.6-1.7`
- R2/KGE very low
- Some runs can get reasonable overall volume/PBIAS, but the monthly/seasonal timing does not match the proxy hydrograph well

My question is: in this situation, is it appropriate to keep trying to calibrate the monthly hydrograph to the scaled proxy discharge, or should I treat the proxy discharge as a soft hydrologic constraint instead?

Would a more defensible strategy be to calibrate/constrain the model using hydrologic signatures such as:

- annual water balance/PBIAS
- seasonal mean flows
- monthly climatology
- flow-duration curve percentiles
- physically realistic parameter ranges

Rather than requiring good monthly R2/KGE against the proxy station?

Also, for a lake watershed modeled with the lake as a reservoir, is `FLOW_INcms` from `output.rsv` the correct variable to calibrate if the goal is watershed loading into the lake, rather than reservoir outflow?

Any suggestions on defensible calibration strategies for an ungauged or proxy-gauged lake watershed would be appreciated.